Lesson 1: How the AI actually reads you
Stayanta's AI is not a single chatbot. It is a small team of specialists — we call them bees — each responsible for one job: pricing, housekeeping dispatch, marketing, reviews, complaints, and so on. When you give an instruction, it is routed to the right specialist. Think of them as a fast, literal, tireless new employee who always writes down exactly what they did.
Getting good results is mostly about giving good instructions. This lesson covers the three things you need to trust the system before you start directing it.
1. Co-Pilot vs Auto-Pilot
- Co-Pilot is the default and where you will live for a long time. The AI does the thinking — reads demand, drafts a reply, works out a rate — then hands you a recommendation card. Nothing happens to your hotel until you press Approve, Edit, or Reject.
- Auto-Pilot is earned, not switched on. A single task stream (say, pricing) can graduate to acting on its own only after it has proven itself over a long track record, and even then it is invite-only and reversible. You will learn exactly how graduation works in Lesson 7.
The mental model: the AI is always the one proposing; in Co-Pilot you decide, in Auto-Pilot the AI decides within limits you already approved and logs every move.
2. The AI can only see your hotel
Every bee is walled inside your hotel's data. It cannot see another hotel on the platform, and it will not benchmark you against, quote, or undercut a named competitor — because it genuinely has no access to that data. This is a hard boundary, not a setting.
Two practical consequences:
- Don't ask a bee to "match the Marriott down the road" — it can't see them. Give it your numbers instead (your floor, your occupancy, your event calendar).
- Anything the AI tells you is grounded in your own hotel's history. That is why the more you use it, the sharper it gets.
3. Everything is logged, and nothing is one-way
Two safety nets sit under every action:
- The Action Ledger records every AI decision — which bee, what it did, when, and why — as a permanent, append-only trail. You can always answer "why did this happen?" (Appendix section 3 shows you how to read it.)
- The Kill Switch is your emergency brake. Flip it and all background AI activity freezes instantly — no rate pushes, no campaigns, no auto-dispatch — and the system drops back to a plain manual PMS until you turn it off. You never lose control by trying something.
Because it is all logged and reversible, you should feel free to experiment. The cost of a less-than-perfect instruction is one rejected card, not a bad night's revenue.